Faculty-student team evaluates storm surge prediction methods using artificial intelligence Meet Sambit Bhattacharya, Ph.D., Grace Vincent, and Raymond Kimble

Along the coast, storm surge from a hurricane is often the greatest threat to life and property. Large death tolls have resulted from the rise of the ocean associated with many of the major hurricanes that have made landfall.  

During the summer of 2020, Sambit Bhattacharya, Ph.D. and a professor of computer science in the computer science and engineering department at Fayetteville State University (FSU), along with FSU computer science students Grace Vincent and Raymond Kimble, examined the problem of storm surge prediction methods in the U.S. Department of Homeland Security (DHS) Summer Research Team (SRT) Program for Minority Serving Institutions (MSI).

bhattacharya-vincent-kimble.jpg

Sambit Bhattacharya, Ph.D., left, Grace Vincent, center and Raymond Kimble, right, conducted research on storm surge prediction methods during the summer of 2020 while participating in the U.S. Department of Homeland Security Summer Research Team Program for Minority Serving Institutions. (Credit: Left, Fayetteville State University, Center, EiPic Productions LLC, Right, Raymond Kimble)

The SRT Program is designed to increase scientific leadership at Minority Serving Institutions in DHS research areas. The program, administered through the U.S. Department of Energy’s (DOE) Oak Ridge Institute for Science and Education (ORISE), provides faculty and student research teams the opportunity to conduct research at university-based DHS Centers of Excellence. 

Throughout their 10-week appointment, Bhattacharya, Vincent and Kimble contributed research at the Coastal Resilience Center (CRC) under the mentorship of Dr. Brian Blanton, a coastal oceanographer and Director of Earth Data Sciences at the Renaissance Computing Institute (RENCI) at the University of North Carolina (UNC), Chapel Hill.

The CRC, a DHS Center of Excellence led by UNC, Chapel Hill, conducts research to enhance the resilience of the nation’s people, infrastructure, economies and the natural environment to the impacts of coastal hazards, such as floods and hurricanes.

“My students and I (were assigned to Dr. Blanton’s team) to study the problem of storm surge prediction and how to develop better prediction methods using the tools of artificial intelligence,” Bhattacharya said. “We (read) scientific literature, designed algorithms, implemented computer programs and ran those programs on experimental data to generate predictions. The predictions were compared with known measurements when possible to understand the accuracy of the methods and to find ways of improving them.”

Because most currently available prediction methods are purely physics-based calculations that are computationally expensive and time consuming, the team’s research purpose was to use observations of winds and water levels to create a hybrid physical machine learning model. “This would provide more accurate predictions of storm surges so that DHS can better protect coastlines from storm surge impacts and be more prepared with their relief efforts,” Vincent said.

One of the most important items the team researched was developing a vast common domain knowledge on the meteorological forces that generate storm surges. Bhattacharya’s team was able to analyze the most prominent inputs and develop a physics-based multi-layer perceptron, a type of computer model devised to simulate the ability of the brain to recognize and discriminate.

“Since this is a challenging problem, we were able to take small but meaningful steps during the program towards our goal of solving the problem,” Bhattacharya said.

Bhattacharya, Vincent and Kimble all agreed that the SRT Program had an impact on their current scientific endeavors. Through a follow-on grant proposal based on their research at CRC, and submitted to DHS by Bhattacharya, they received one year of follow-on funding from DHS to continue their efforts to improve storm surge prediction methods in an ongoing collaboration with Dr. Blanton.

“The impact of the SRT experience was in creating a new area of interest in my research activities,” Bhattacharya said. “I expect that this seed award will help my team develop a larger research program in collaboration with our host institution.”

Kimble and Vincent each had their own takeaways from their SRT collaboration and some advice for future participants interested in summer research opportunities.

“It was a great learning experience (using data research) at CRC and using it to establish a physics-based neural network that can accurately predict storm surges from hurricanes and tropical storms,” Kimble said. “My advice about STEM-focused internships is to do it with the willingness to learn about problem-solving, learn from mistakes and be open-minded to new ideas.”

Vincent said, “My time with ORISE has opened many doors for me through networking and experience gained. This experience showed me the relevance of the work I am doing and has pointed me towards the pursuit of future federal agency internships or research projects. I believe this experience will be an influential step into my future career in machine learning.”

Bhattacharya strongly recommended the ORISE program to students, saying, “I believe this program can be a life changing experience for STEM students since they will gain knowledge of practical, important research problems and they will be able to practice their problem-solving skills by developing solutions to these problems.”

A parting piece of advice from Bhattacharya, “Do not be discouraged by any initial setbacks like not being able to understand the research problem or background literature on the problem. Research advisors at host locations understand that helping students to develop an understanding of the problems, along with the research mindset, is important during the short duration experience.”

The DHS SRT MSI Program is funded by DHS and administered through the U.S. Department of Energy’s (DOE) Oak Ridge Institute for Science and Education (ORISE). ORISE is managed for DOE by Oak Ridge Associated Universities.